You are viewing the course site for a past offering of this course. The current offering may be found here.
Lecture 23: How GPUs work (23)
ellenluo

I was curious about what assembly language is used for GPUs since I don't think x86 (and other common languages) allow for operations like 'sample' and 'clmp'.

shannonsh

You are correct in that GPU instructions are different from CPUs. You might've heard about "Nvidia CUDA" a lot. Turns out CUDA is an API Nvidia created so programmers can directly access the Nvidia GPU's instruction set. Since GPUs have no standardized instruction set, different GPUs made by the same manufacturer could have different instruction sets. For more information see here: https://en.wikipedia.org/wiki/CUDA

hershg

OpenCL is an open source alternative to CUDA for working w/ GPU/different accelerator instruction sets with a higher level C-type language. Popular among lots of non-Nvidia companies (like Intel, Apple) who don't want full market domination of Nvidia/CUDA workflow

You must be enrolled in the course to comment